Quick check
Guess first, no penalty. Your AI forecast shows 1,000 units a week for four weeks. A promotion runs in week 3, and the last two like it doubled sales. What do you do?
A demand forecast predicts how much you'll sell or ship. AI forecasts learn from past data, so they're good at normal weeks and blind to anything new: a promotion, a moved holiday, a strike, a new customer, a lost one.
A flat line across a week you know is different is the warning sign. The model isn't broken. It just doesn't know.
So before you plan from any forecast, ask one question for each coming week: what do I know that the data doesn't?
Worked example
Before and after
The AI forecast, then the checked one with its note.
AI forecast
Week 1: 1,000. Week 2: 1,000. Week 3: 1,000. Week 4: 1,000.
The note matters as much as the number. The next planner sees what changed, who changed it and why.
To see what a missed promotion does to stock, try the Safety stock lab.
Fill the gaps
Fill the gaps in your forecast check.
Before I plan from a forecast
For each week I ask: . If I change a number, I .
Try it
Your steps, about 2 minutes
- 1.Add the forecast question to your AI check card, a one-page note you'll build in this module. Then look at next month and list every week the data can't know about.2 min